3 * Linux ethernet bridge
6 * Lennert Buytenhek <buytenh@gnu.org>
7 * Bart De Schuymer (maintainer) <bdschuym@pandora.be>
10 * Apr 29 2003: physdev module support (bdschuym)
11 * Jun 19 2003: let arptables see bridged ARP traffic (bdschuym)
12 * Oct 06 2003: filter encapsulated IP/ARP VLAN traffic on untagged bridge
14 * Sep 01 2004: add IPv6 filtering (bdschuym)
16 * This program is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU General Public License
18 * as published by the Free Software Foundation; either version
19 * 2 of the License, or (at your option) any later version.
21 * Lennert dedicates this file to Kerstin Wurdinger.
24 #include <linux/module.h>
25 #include <linux/kernel.h>
26 #include <linux/slab.h>
28 #include <linux/netdevice.h>
29 #include <linux/skbuff.h>
30 #include <linux/if_arp.h>
31 #include <linux/if_ether.h>
32 #include <linux/if_vlan.h>
33 #include <linux/if_pppox.h>
34 #include <linux/ppp_defs.h>
35 #include <linux/netfilter_bridge.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <linux/netfilter_ipv6.h>
38 #include <linux/netfilter_arp.h>
39 #include <linux/in_route.h>
40 #include <linux/inetdevice.h>
44 #include <net/route.h>
46 #include <asm/uaccess.h>
47 #include "br_private.h"
49 #include <linux/sysctl.h>
52 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
53 (skb->nf_bridge->data))->daddr.ipv4)
54 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
55 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
58 static struct ctl_table_header
*brnf_sysctl_header
;
59 static int brnf_call_iptables __read_mostly
= 1;
60 static int brnf_call_ip6tables __read_mostly
= 1;
61 static int brnf_call_arptables __read_mostly
= 1;
62 static int brnf_filter_vlan_tagged __read_mostly
= 0;
63 static int brnf_filter_pppoe_tagged __read_mostly
= 0;
65 #define brnf_filter_vlan_tagged 0
66 #define brnf_filter_pppoe_tagged 0
69 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
71 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
74 #define IS_VLAN_IP(skb) \
75 (skb->protocol == htons(ETH_P_8021Q) && \
76 vlan_proto(skb) == htons(ETH_P_IP) && \
77 brnf_filter_vlan_tagged)
79 #define IS_VLAN_IPV6(skb) \
80 (skb->protocol == htons(ETH_P_8021Q) && \
81 vlan_proto(skb) == htons(ETH_P_IPV6) &&\
82 brnf_filter_vlan_tagged)
84 #define IS_VLAN_ARP(skb) \
85 (skb->protocol == htons(ETH_P_8021Q) && \
86 vlan_proto(skb) == htons(ETH_P_ARP) && \
87 brnf_filter_vlan_tagged)
89 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
91 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
92 sizeof(struct pppoe_hdr
)));
95 #define IS_PPPOE_IP(skb) \
96 (skb->protocol == htons(ETH_P_PPP_SES) && \
97 pppoe_proto(skb) == htons(PPP_IP) && \
98 brnf_filter_pppoe_tagged)
100 #define IS_PPPOE_IPV6(skb) \
101 (skb->protocol == htons(ETH_P_PPP_SES) && \
102 pppoe_proto(skb) == htons(PPP_IPV6) && \
103 brnf_filter_pppoe_tagged)
105 static void fake_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
109 static struct dst_ops fake_dst_ops
= {
111 .protocol
= cpu_to_be16(ETH_P_IP
),
112 .update_pmtu
= fake_update_pmtu
,
113 .entries
= ATOMIC_INIT(0),
117 * Initialize bogus route table used to keep netfilter happy.
118 * Currently, we fill in the PMTU entry because netfilter
119 * refragmentation needs it, and the rt_flags entry because
120 * ipt_REJECT needs it. Future netfilter modules might
121 * require us to fill additional fields.
123 void br_netfilter_rtable_init(struct net_bridge
*br
)
125 struct rtable
*rt
= &br
->fake_rtable
;
127 atomic_set(&rt
->u
.dst
.__refcnt
, 1);
128 rt
->u
.dst
.dev
= br
->dev
;
129 rt
->u
.dst
.path
= &rt
->u
.dst
;
130 rt
->u
.dst
.metrics
[RTAX_MTU
- 1] = 1500;
131 rt
->u
.dst
.flags
= DST_NOXFRM
;
132 rt
->u
.dst
.ops
= &fake_dst_ops
;
135 static inline struct rtable
*bridge_parent_rtable(const struct net_device
*dev
)
137 struct net_bridge_port
*port
= rcu_dereference(dev
->br_port
);
139 return port
? &port
->br
->fake_rtable
: NULL
;
142 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
144 struct net_bridge_port
*port
= rcu_dereference(dev
->br_port
);
146 return port
? port
->br
->dev
: NULL
;
149 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
151 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
152 if (likely(skb
->nf_bridge
))
153 atomic_set(&(skb
->nf_bridge
->use
), 1);
155 return skb
->nf_bridge
;
158 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
160 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
162 if (atomic_read(&nf_bridge
->use
) > 1) {
163 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
166 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
167 atomic_set(&tmp
->use
, 1);
168 nf_bridge_put(nf_bridge
);
175 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
177 unsigned int len
= nf_bridge_encap_header_len(skb
);
180 skb
->network_header
-= len
;
183 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
185 unsigned int len
= nf_bridge_encap_header_len(skb
);
188 skb
->network_header
+= len
;
191 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
193 unsigned int len
= nf_bridge_encap_header_len(skb
);
195 skb_pull_rcsum(skb
, len
);
196 skb
->network_header
+= len
;
199 static inline void nf_bridge_save_header(struct sk_buff
*skb
)
201 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
203 skb_copy_from_linear_data_offset(skb
, -header_size
,
204 skb
->nf_bridge
->data
, header_size
);
208 * When forwarding bridge frames, we save a copy of the original
209 * header before processing.
211 int nf_bridge_copy_header(struct sk_buff
*skb
)
214 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
216 err
= skb_cow_head(skb
, header_size
);
220 skb_copy_to_linear_data_offset(skb
, -header_size
,
221 skb
->nf_bridge
->data
, header_size
);
222 __skb_push(skb
, nf_bridge_encap_header_len(skb
));
226 /* PF_BRIDGE/PRE_ROUTING *********************************************/
227 /* Undo the changes made for ip6tables PREROUTING and continue the
228 * bridge PRE_ROUTING hook. */
229 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
231 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
234 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
235 skb
->pkt_type
= PACKET_OTHERHOST
;
236 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
238 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
240 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
245 dst_hold(&rt
->u
.dst
);
246 skb_dst_set(skb
, &rt
->u
.dst
);
248 skb
->dev
= nf_bridge
->physindev
;
249 nf_bridge_push_encap_header(skb
);
250 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
251 br_handle_frame_finish
, 1);
256 static void __br_dnat_complain(void)
258 static unsigned long last_complaint
;
260 if (jiffies
- last_complaint
>= 5 * HZ
) {
261 printk(KERN_WARNING
"Performing cross-bridge DNAT requires IP "
262 "forwarding to be enabled\n");
263 last_complaint
= jiffies
;
267 /* This requires some explaining. If DNAT has taken place,
268 * we will need to fix up the destination Ethernet address,
269 * and this is a tricky process.
271 * There are two cases to consider:
272 * 1. The packet was DNAT'ed to a device in the same bridge
273 * port group as it was received on. We can still bridge
275 * 2. The packet was DNAT'ed to a different device, either
276 * a non-bridged device or another bridge port group.
277 * The packet will need to be routed.
279 * The correct way of distinguishing between these two cases is to
280 * call ip_route_input() and to look at skb->dst->dev, which is
281 * changed to the destination device if ip_route_input() succeeds.
283 * Let us first consider the case that ip_route_input() succeeds:
285 * If skb->dst->dev equals the logical bridge device the packet
286 * came in on, we can consider this bridging. The packet is passed
287 * through the neighbour output function to build a new destination
288 * MAC address, which will make the packet enter br_nf_local_out()
289 * not much later. In that function it is assured that the iptables
290 * FORWARD chain is traversed for the packet.
292 * Otherwise, the packet is considered to be routed and we just
293 * change the destination MAC address so that the packet will
294 * later be passed up to the IP stack to be routed. For a redirected
295 * packet, ip_route_input() will give back the localhost as output device,
296 * which differs from the bridge device.
298 * Let us now consider the case that ip_route_input() fails:
300 * This can be because the destination address is martian, in which case
301 * the packet will be dropped.
302 * After a "echo '0' > /proc/sys/net/ipv4/ip_forward" ip_route_input()
303 * will fail, while __ip_route_output_key() will return success. The source
304 * address for __ip_route_output_key() is set to zero, so __ip_route_output_key
305 * thinks we're handling a locally generated packet and won't care
306 * if IP forwarding is allowed. We send a warning message to the users's
307 * log telling her to put IP forwarding on.
309 * ip_route_input() will also fail if there is no route available.
310 * In that case we just drop the packet.
312 * --Lennert, 20020411
313 * --Bart, 20020416 (updated)
314 * --Bart, 20021007 (updated)
315 * --Bart, 20062711 (updated) */
316 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
318 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
319 skb
->pkt_type
= PACKET_HOST
;
320 skb
->nf_bridge
->mask
|= BRNF_PKT_TYPE
;
322 skb
->nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
324 skb
->dev
= bridge_parent(skb
->dev
);
326 struct dst_entry
*dst
= skb_dst(skb
);
328 nf_bridge_pull_encap_header(skb
);
331 return neigh_hh_output(dst
->hh
, skb
);
332 else if (dst
->neighbour
)
333 return dst
->neighbour
->output(skb
);
339 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
341 struct net_device
*dev
= skb
->dev
;
342 struct iphdr
*iph
= ip_hdr(skb
);
343 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
347 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
348 skb
->pkt_type
= PACKET_OTHERHOST
;
349 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
351 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
352 if (dnat_took_place(skb
)) {
353 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
359 .tos
= RT_TOS(iph
->tos
) },
363 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
365 /* If err equals -EHOSTUNREACH the error is due to a
366 * martian destination or due to the fact that
367 * forwarding is disabled. For most martian packets,
368 * ip_route_output_key() will fail. It won't fail for 2 types of
369 * martian destinations: loopback destinations and destination
370 * 0.0.0.0. In both cases the packet will be dropped because the
371 * destination is the loopback device and not the bridge. */
372 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
375 if (!ip_route_output_key(dev_net(dev
), &rt
, &fl
)) {
376 /* - Bridged-and-DNAT'ed traffic doesn't
377 * require ip_forwarding. */
378 if (((struct dst_entry
*)rt
)->dev
== dev
) {
379 skb_dst_set(skb
, (struct dst_entry
*)rt
);
382 /* we are sure that forwarding is disabled, so printing
383 * this message is no problem. Note that the packet could
384 * still have a martian destination address, in which case
385 * the packet could be dropped even if forwarding were enabled */
386 __br_dnat_complain();
387 dst_release((struct dst_entry
*)rt
);
393 if (skb_dst(skb
)->dev
== dev
) {
395 /* Tell br_nf_local_out this is a
397 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
398 skb
->dev
= nf_bridge
->physindev
;
399 nf_bridge_push_encap_header(skb
);
400 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_PRE_ROUTING
,
402 br_nf_pre_routing_finish_bridge
,
406 memcpy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
, ETH_ALEN
);
407 skb
->pkt_type
= PACKET_HOST
;
410 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
415 dst_hold(&rt
->u
.dst
);
416 skb_dst_set(skb
, &rt
->u
.dst
);
419 skb
->dev
= nf_bridge
->physindev
;
420 nf_bridge_push_encap_header(skb
);
421 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
422 br_handle_frame_finish
, 1);
427 /* Some common code for IPv4/IPv6 */
428 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
430 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
432 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
433 skb
->pkt_type
= PACKET_HOST
;
434 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
437 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
438 nf_bridge
->physindev
= skb
->dev
;
439 skb
->dev
= bridge_parent(skb
->dev
);
444 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
445 static int check_hbh_len(struct sk_buff
*skb
)
447 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
449 const unsigned char *nh
= skb_network_header(skb
);
451 int len
= (raw
[1] + 1) << 3;
453 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
460 int optlen
= nh
[off
+ 1] + 2;
471 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
473 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
474 if (pkt_len
<= IPV6_MAXPLEN
||
475 ipv6_hdr(skb
)->payload_len
)
477 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
479 if (pskb_trim_rcsum(skb
,
480 pkt_len
+ sizeof(struct ipv6hdr
)))
482 nh
= skb_network_header(skb
);
499 /* Replicate the checks that IPv6 does on packet reception and pass the packet
500 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
501 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook
,
503 const struct net_device
*in
,
504 const struct net_device
*out
,
505 int (*okfn
)(struct sk_buff
*))
510 if (skb
->len
< sizeof(struct ipv6hdr
))
513 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
518 if (hdr
->version
!= 6)
521 pkt_len
= ntohs(hdr
->payload_len
);
523 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
524 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
526 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
529 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
532 nf_bridge_put(skb
->nf_bridge
);
533 if (!nf_bridge_alloc(skb
))
535 if (!setup_pre_routing(skb
))
538 NF_HOOK(PF_INET6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
539 br_nf_pre_routing_finish_ipv6
);
547 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
548 * Replicate the checks that IPv4 does on packet reception.
549 * Set skb->dev to the bridge device (i.e. parent of the
550 * receiving device) to make netfilter happy, the REDIRECT
551 * target in particular. Save the original destination IP
552 * address to be able to detect DNAT afterwards. */
553 static unsigned int br_nf_pre_routing(unsigned int hook
, struct sk_buff
*skb
,
554 const struct net_device
*in
,
555 const struct net_device
*out
,
556 int (*okfn
)(struct sk_buff
*))
559 __u32 len
= nf_bridge_encap_header_len(skb
);
561 if (unlikely(!pskb_may_pull(skb
, len
)))
564 if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
565 IS_PPPOE_IPV6(skb
)) {
567 if (!brnf_call_ip6tables
)
570 nf_bridge_pull_encap_header_rcsum(skb
);
571 return br_nf_pre_routing_ipv6(hook
, skb
, in
, out
, okfn
);
574 if (!brnf_call_iptables
)
578 if (skb
->protocol
!= htons(ETH_P_IP
) && !IS_VLAN_IP(skb
) &&
582 nf_bridge_pull_encap_header_rcsum(skb
);
584 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
588 if (iph
->ihl
< 5 || iph
->version
!= 4)
591 if (!pskb_may_pull(skb
, 4 * iph
->ihl
))
595 if (ip_fast_csum((__u8
*) iph
, iph
->ihl
) != 0)
598 len
= ntohs(iph
->tot_len
);
599 if (skb
->len
< len
|| len
< 4 * iph
->ihl
)
602 pskb_trim_rcsum(skb
, len
);
604 nf_bridge_put(skb
->nf_bridge
);
605 if (!nf_bridge_alloc(skb
))
607 if (!setup_pre_routing(skb
))
609 store_orig_dstaddr(skb
);
611 NF_HOOK(PF_INET
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
612 br_nf_pre_routing_finish
);
617 // IP_INC_STATS_BH(IpInHdrErrors);
623 /* PF_BRIDGE/LOCAL_IN ************************************************/
624 /* The packet is locally destined, which requires a real
625 * dst_entry, so detach the fake one. On the way up, the
626 * packet would pass through PRE_ROUTING again (which already
627 * took place when the packet entered the bridge), but we
628 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
629 * prevent this from happening. */
630 static unsigned int br_nf_local_in(unsigned int hook
, struct sk_buff
*skb
,
631 const struct net_device
*in
,
632 const struct net_device
*out
,
633 int (*okfn
)(struct sk_buff
*))
635 struct rtable
*rt
= skb_rtable(skb
);
637 if (rt
&& rt
== bridge_parent_rtable(in
))
643 /* PF_BRIDGE/FORWARD *************************************************/
644 static int br_nf_forward_finish(struct sk_buff
*skb
)
646 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
647 struct net_device
*in
;
649 if (skb
->protocol
!= htons(ETH_P_ARP
) && !IS_VLAN_ARP(skb
)) {
650 in
= nf_bridge
->physindev
;
651 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
652 skb
->pkt_type
= PACKET_OTHERHOST
;
653 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
656 in
= *((struct net_device
**)(skb
->cb
));
658 nf_bridge_push_encap_header(skb
);
659 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
660 skb
->dev
, br_forward_finish
, 1);
664 /* This is the 'purely bridged' case. For IP, we pass the packet to
665 * netfilter with indev and outdev set to the bridge device,
666 * but we are still able to filter on the 'real' indev/outdev
667 * because of the physdev module. For ARP, indev and outdev are the
669 static unsigned int br_nf_forward_ip(unsigned int hook
, struct sk_buff
*skb
,
670 const struct net_device
*in
,
671 const struct net_device
*out
,
672 int (*okfn
)(struct sk_buff
*))
674 struct nf_bridge_info
*nf_bridge
;
675 struct net_device
*parent
;
681 /* Need exclusive nf_bridge_info since we might have multiple
682 * different physoutdevs. */
683 if (!nf_bridge_unshare(skb
))
686 parent
= bridge_parent(out
);
690 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
693 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
699 nf_bridge_pull_encap_header(skb
);
701 nf_bridge
= skb
->nf_bridge
;
702 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
703 skb
->pkt_type
= PACKET_HOST
;
704 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
707 /* The physdev module checks on this */
708 nf_bridge
->mask
|= BRNF_BRIDGED
;
709 nf_bridge
->physoutdev
= skb
->dev
;
711 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, bridge_parent(in
), parent
,
712 br_nf_forward_finish
);
717 static unsigned int br_nf_forward_arp(unsigned int hook
, struct sk_buff
*skb
,
718 const struct net_device
*in
,
719 const struct net_device
*out
,
720 int (*okfn
)(struct sk_buff
*))
722 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
725 if (!brnf_call_arptables
)
729 if (skb
->protocol
!= htons(ETH_P_ARP
)) {
730 if (!IS_VLAN_ARP(skb
))
732 nf_bridge_pull_encap_header(skb
);
735 if (arp_hdr(skb
)->ar_pln
!= 4) {
736 if (IS_VLAN_ARP(skb
))
737 nf_bridge_push_encap_header(skb
);
740 *d
= (struct net_device
*)in
;
741 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
742 (struct net_device
*)out
, br_nf_forward_finish
);
747 /* PF_BRIDGE/LOCAL_OUT ***********************************************
749 * This function sees both locally originated IP packets and forwarded
750 * IP packets (in both cases the destination device is a bridge
751 * device). It also sees bridged-and-DNAT'ed packets.
753 * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
754 * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
755 * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
756 * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
759 static unsigned int br_nf_local_out(unsigned int hook
, struct sk_buff
*skb
,
760 const struct net_device
*in
,
761 const struct net_device
*out
,
762 int (*okfn
)(struct sk_buff
*))
764 struct net_device
*realindev
;
765 struct nf_bridge_info
*nf_bridge
;
770 /* Need exclusive nf_bridge_info since we might have multiple
771 * different physoutdevs. */
772 if (!nf_bridge_unshare(skb
))
775 nf_bridge
= skb
->nf_bridge
;
776 if (!(nf_bridge
->mask
& BRNF_BRIDGED_DNAT
))
779 /* Bridged, take PF_BRIDGE/FORWARD.
780 * (see big note in front of br_nf_pre_routing_finish) */
781 nf_bridge
->physoutdev
= skb
->dev
;
782 realindev
= nf_bridge
->physindev
;
784 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
785 skb
->pkt_type
= PACKET_OTHERHOST
;
786 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
788 nf_bridge_push_encap_header(skb
);
790 NF_HOOK(PF_BRIDGE
, NF_BR_FORWARD
, skb
, realindev
, skb
->dev
,
795 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
796 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
798 if (skb
->nfct
!= NULL
&&
799 (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
)) &&
800 skb
->len
> skb
->dev
->mtu
&&
802 return ip_fragment(skb
, br_dev_queue_push_xmit
);
804 return br_dev_queue_push_xmit(skb
);
807 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
809 return br_dev_queue_push_xmit(skb
);
813 /* PF_BRIDGE/POST_ROUTING ********************************************/
814 static unsigned int br_nf_post_routing(unsigned int hook
, struct sk_buff
*skb
,
815 const struct net_device
*in
,
816 const struct net_device
*out
,
817 int (*okfn
)(struct sk_buff
*))
819 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
820 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
823 #ifdef CONFIG_NETFILTER_DEBUG
824 /* Be very paranoid. This probably won't happen anymore, but let's
825 * keep the check just to be sure... */
826 if (skb_mac_header(skb
) < skb
->head
||
827 skb_mac_header(skb
) + ETH_HLEN
> skb
->data
) {
828 printk(KERN_CRIT
"br_netfilter: Argh!! br_nf_post_routing: "
829 "bad mac.raw pointer.\n");
837 if (!(nf_bridge
->mask
& (BRNF_BRIDGED
| BRNF_BRIDGED_DNAT
)))
843 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
846 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
852 #ifdef CONFIG_NETFILTER_DEBUG
853 if (skb_dst(skb
) == NULL
) {
854 printk(KERN_INFO
"br_netfilter post_routing: skb->dst == NULL\n");
859 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
860 * about the value of skb->pkt_type. */
861 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
862 skb
->pkt_type
= PACKET_HOST
;
863 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
866 nf_bridge_pull_encap_header(skb
);
867 nf_bridge_save_header(skb
);
869 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
870 br_nf_dev_queue_xmit
);
874 #ifdef CONFIG_NETFILTER_DEBUG
876 if (skb
->dev
!= NULL
) {
877 printk("[%s]", skb
->dev
->name
);
879 printk("[%s]", realoutdev
->name
);
881 printk(" head:%p, raw:%p, data:%p\n", skb
->head
, skb_mac_header(skb
),
888 /* IP/SABOTAGE *****************************************************/
889 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
890 * for the second time. */
891 static unsigned int ip_sabotage_in(unsigned int hook
, struct sk_buff
*skb
,
892 const struct net_device
*in
,
893 const struct net_device
*out
,
894 int (*okfn
)(struct sk_buff
*))
896 if (skb
->nf_bridge
&&
897 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
904 /* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
905 * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
906 * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
907 * ip_refrag() can return NF_STOLEN. */
908 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
910 .hook
= br_nf_pre_routing
,
911 .owner
= THIS_MODULE
,
913 .hooknum
= NF_BR_PRE_ROUTING
,
914 .priority
= NF_BR_PRI_BRNF
,
917 .hook
= br_nf_local_in
,
918 .owner
= THIS_MODULE
,
920 .hooknum
= NF_BR_LOCAL_IN
,
921 .priority
= NF_BR_PRI_BRNF
,
924 .hook
= br_nf_forward_ip
,
925 .owner
= THIS_MODULE
,
927 .hooknum
= NF_BR_FORWARD
,
928 .priority
= NF_BR_PRI_BRNF
- 1,
931 .hook
= br_nf_forward_arp
,
932 .owner
= THIS_MODULE
,
934 .hooknum
= NF_BR_FORWARD
,
935 .priority
= NF_BR_PRI_BRNF
,
938 .hook
= br_nf_local_out
,
939 .owner
= THIS_MODULE
,
941 .hooknum
= NF_BR_LOCAL_OUT
,
942 .priority
= NF_BR_PRI_FIRST
,
945 .hook
= br_nf_post_routing
,
946 .owner
= THIS_MODULE
,
948 .hooknum
= NF_BR_POST_ROUTING
,
949 .priority
= NF_BR_PRI_LAST
,
952 .hook
= ip_sabotage_in
,
953 .owner
= THIS_MODULE
,
955 .hooknum
= NF_INET_PRE_ROUTING
,
956 .priority
= NF_IP_PRI_FIRST
,
959 .hook
= ip_sabotage_in
,
960 .owner
= THIS_MODULE
,
962 .hooknum
= NF_INET_PRE_ROUTING
,
963 .priority
= NF_IP6_PRI_FIRST
,
969 int brnf_sysctl_call_tables(ctl_table
* ctl
, int write
,
970 void __user
* buffer
, size_t * lenp
, loff_t
* ppos
)
974 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
976 if (write
&& *(int *)(ctl
->data
))
977 *(int *)(ctl
->data
) = 1;
981 static ctl_table brnf_table
[] = {
983 .procname
= "bridge-nf-call-arptables",
984 .data
= &brnf_call_arptables
,
985 .maxlen
= sizeof(int),
987 .proc_handler
= brnf_sysctl_call_tables
,
990 .procname
= "bridge-nf-call-iptables",
991 .data
= &brnf_call_iptables
,
992 .maxlen
= sizeof(int),
994 .proc_handler
= brnf_sysctl_call_tables
,
997 .procname
= "bridge-nf-call-ip6tables",
998 .data
= &brnf_call_ip6tables
,
999 .maxlen
= sizeof(int),
1001 .proc_handler
= brnf_sysctl_call_tables
,
1004 .procname
= "bridge-nf-filter-vlan-tagged",
1005 .data
= &brnf_filter_vlan_tagged
,
1006 .maxlen
= sizeof(int),
1008 .proc_handler
= brnf_sysctl_call_tables
,
1011 .procname
= "bridge-nf-filter-pppoe-tagged",
1012 .data
= &brnf_filter_pppoe_tagged
,
1013 .maxlen
= sizeof(int),
1015 .proc_handler
= brnf_sysctl_call_tables
,
1020 static struct ctl_path brnf_path
[] = {
1021 { .procname
= "net", },
1022 { .procname
= "bridge", },
1027 int __init
br_netfilter_init(void)
1031 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1034 #ifdef CONFIG_SYSCTL
1035 brnf_sysctl_header
= register_sysctl_paths(brnf_path
, brnf_table
);
1036 if (brnf_sysctl_header
== NULL
) {
1038 "br_netfilter: can't register to sysctl.\n");
1039 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1043 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1047 void br_netfilter_fini(void)
1049 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1050 #ifdef CONFIG_SYSCTL
1051 unregister_sysctl_table(brnf_sysctl_header
);